Abstract: Gas turbine combined cycle unit is an important equipment involved in peak and valley regulation and frequency regulation of power grid, and its operation performance is directly related to the security of the power grid. Aiming at improving the reliability of power grid, a hybrid-form adaptive selection algorithm model (HASAM) is proposed in this paper. First, combined with the configuration logic of the manufacturer's control strategy and partial mutual information, variables related to the gas turbine calculation power are selected as modeling objects. Then, the modeling sub-model is constructed by the quantum particle swarm (QPSO) identification model, the deep neural network (DNN) algorithm and the simple recurrent neural network (ELMAN) algorithm. Finally, according to different working conditions, the DBN classification model optimized by QPSO adaptively selects the sub-model to predict the calculated power of the gas turbine. The results based on 9E gas turbine operating data illustrate that the proposed algorithm could accurately predict the calculated power of gas turbine.